Methodologies as extracted from the
DATA REPORT
NORTH ATLANTIC BLOOM EXPERIMENT
APRIL - JULY 1989
Prepared by
Lt. Raymond Slagle
George Heimerdinger
NODC/U.S. JGOFS Data Management Office
Woods Hole Oceanographic Institution
Woods Hole, MA 02543
November 1990
TABLE OF CONTENTS
2. METHODS CTD DATA
2.1 CTD PROFILES
(Williams, SIO)
- 2.1.1 CTD Laboratory Calibrations
- 2.1.2 Data Acquisition and Display
- 2.1.3 Post-Acquisition and Post-Cruise Processing
- 2.1.3.1 Pressure, Temperature, and Salinity
- 2.1.3.2 Disolved Oxygen Data
- 2.1.3.3 Additional Processing
2.2 BEAM ATTENUATION
(Gardner, TAMU)
2.3 CTD PROFILES
(Broenkow, MLML)
- 2.3.1 Oxygen
- 2.3.2 Beam attenuation
- 2.3.3 Fluorescence
2.1 CTD PROFILES (Williams, SIO)
All SIO CTD profiles were taken with a CTD constructed by personnel of the Oceanographic
Data Facility (ODF) at SIO, using NBIS Mark IIIB circuit cards, some of which
have been modified, together with some cards designed and produced by ODF. Pressure,
temperature, conductivity, and oxygen sensors were standard sensors as used
in the factory-produced Mark III CTD units. The pressure and oxygen circuit
design was not original, however, but was modified to suit ODF data handling
and processing philosophies and techniques. The CTD data included one 16-bit
and one 8-bit multiplexed channel for the transmission of non-standard parameters,
including elapsed time, transmissometer (see Section 2.2), several internal
voltages for diagnostic purposes, and rosette-trip confirmation data.
2.1.1 CTD Laboratory Calibrations
Before and after the North Atlantic Bloom Experiment, the CTD pressure and temperature
channels were calibrated in the ODF calibration facility. While deep sea reversing
thermometers provide some useful information in the event of a detectable change
at sea, in general the CTD is more precise than thermometers, and accuracy is
enhanced by using high-precision calibration systems ashore. The pressure standard
in use at ODF is a Ruska Model 2400 Piston Gage providing pressures accurate
to better than 0.01%. The CTD pressure transducer, a Paine Instruments Model
211-35 strain gage, was calibrated at 23 points between 0 and 6100 db, both
as pressure was increased and decreased. Its performance was observed at several
different temperatures, and to varying maximum pressures, to note the temperature
coefficient and the change in hysteresis with depth of cast. Additionally, the
transient response of the pressure transducer to temperature changes was observed.
The platinum resistance temperature sensor, Rosemount Model 171BJ, was calibrated
at several points throughout the temperature range encountered during the expedition,
so that temperature could be corrected to within 1 mdeg C. The temperature transfer
standard at the ODF calibration facility, a NBIS Model ATB-1250 automatic resistance
bridge and Rosemount Model 162CE standard platinum thermometer, is calibrated
periodically against several water triple point cells at 0.01 deg C, and a phenoxy-benzene
triple point cell at 26.868 deg C. CTD calibrations are performed in the ODF
calibration bath, in a 350 liter tank equipped with a Tronac control system.
The tank maintains a stability of 0.0002 deg C for as long as is required, and
has internal temperature gradients of no more than 1.5 mdeg/meter. It should
be clearly noted that the temperatures produced by ODF for this program are
on the IPTS-68 temperature scale, rather than the new ITS-90 scale. No laboratory
calibrations were performed for either conductivity or oxygen, as it is necessary
to collect samples at sea to assure reasonable accuracy from these sensors.
2.1.2 Data Acquisition and Display
CTD data was acquired at a sampling rate of 25 Hz. Raw data was recorded on
the audio channel of a video cassette recorder, and in digital form on a hard
disk. Data reduction occurred in realtime, converting the 25 Hz data to a 0.5
second time series. During the reduction, the conductivity and pressure channels
were lagged by a single pole, low-pass filter to match the response of the temperature
sensor, which had a time constant of approximately 500 msec. Individual frames
of data were subjected to a filter limiting absolute values and gradients to
reasonable and possible values. The data were then averaged into half-second
blocks, and the mean and standard deviation of individual points from the mean
were calculated. Editing of the data was done with a two-pass standard deviation
rejection filter. On the first pass, values exceeding 4 standard devitions from
the mean were rejected, and the remaining data were re-averaged. On the second
pass, values exceeding 2 standard deviations were rejected, again followed by
re-averaging. These half-second blocks of data, with pre-cruise calibration
data applied, were displayed on a high-resolution monitor in graphic form during
the cast. During each bottle-trip, about 10 seconds of data were averaged to
represent the pressure and temperature for the bottle data file.
2.1.3 Post-Acquisition and Post-Cruise Processing
Individual water sample salinities were determined on almost all casts, in sufficient
number to closely monitor and correct the EG&G conductivity sensor. After the
first few casts, sufficient information was on hand to provide reasonably good
realtime corrections, and permit the production of usable shipboard data reports.
Dissolved oxygen analyses were performed on many water samples, but no effort
was made to process the CTD oxygen sensor data at sea. Following the cruise,
the CTD pressure and temperature sensors were completely recalibrated. No significant
changes were observed. The pre-cruise and post-cruise calibrations were averaged
and used for all casts. The profiles reported here are primarily those taken
while lowering the rosette to maximum depth, unless excessive noise or other
data problems made the downtrace data unacceptable (a rare occurence in this
expedition). This procedure normally provides the cleanest data, avoiding the
stops and starts associated with the collection of water samples as the rosette
is raised back to the surface. It will be noted that there are occasions where
bottle data and CTD data at a given level are different as a result of the different
times of data and/or sample collection.
2.1.3.1 Pressure, Temperature, and Salinity
Pressures are considered to be accurate to within 2 db. with a precision of
1 db, cast-to-cast. Correction of the raw data was accomplished with a model
of dynamic response of the pressure transducer which takes into consideration
pressure error as a function of pressure and dynamically changing temperature,
as well as the variable hysteresis of this type of sensor as a function of maximum
cast pressure. Temperature accuracy is estimated to be within 0.001 deg C below
10 deg C, and within 0.002 deg C above 10 deg C, with a precision of 0.0005
deg C. The CTD salinities were adjusted to match the bottle salinities unless
there was good reason to suspect a particular cast of bottle salinities. Such
a reason might be found in the close matching of several CTD potential temperature-salinity
diagrams using a constant conductivity correction, when the bottle data suggests
changing the CTD conductivity corrections. Both bottle and CTD data were subjected
to very close scrutiny following the cruise, to avoid changing CTD data to match
bottle data exactly and arbitrarily, when experience demonstrates clearly that
bottle salinities are not infallible. In all cases, the conductivity was corrected
by adjusting an offset term; there was no change in the slope of conductivity
as a simple first-order function of conductivity calculated from bottle salinities
and corrected CTD temperatures and pressures. During this expedition, both the
salinometer and CTD performed well, and the CTD salinity accuracy is estimated
to be the same as the bottle accuracy, within 0.002 psu, with a precision of
0.001 from cast-to-cast.
2.1.3.2 Dissolved Oxygen Data
Dissolved oxygen data were acquired using a Sensormedics (formerly Beckman)
dissolved oxygen sensor. CTD downcast raw oxygen current was extracted from
the corrected pressure-series data at isopycnals corresponding to the upcast
bottle samples. The differences between CTD and bottle oxygens were used on
a station-by-station basis to generate coefficients for a sensor model by applying
a non-linear fitting procedure. The model includes pressure and temperature
effects on sensor membrane permeability. The temperature of the membrane was
calculated via low-pass filter from CTD temperature, and used to compute the
diffusion time constant for the membrane. CTD pressure, temperature and salinity
were lagged to match the O2 response. Oxygen partial pressure was calculated
and converted to dissolved oxygen concentration according to Weiss (1970). The
oxygen sensor used in the NBIS Mark III CTD is not ideally suited for oceanographic
applications in that it consumes both oxygen and itself during the course of
measurements, and is therefore inherently unstable over the long term. It often
requires several seconds in the water before it is wet enough to respond properly;
this is manifested as low oxygen values at the start of some casts. Flow-dependence
problems may occur when the lowering rate varies, as at the cast bottom or during
bottle trips, where depletion of oxygen at the sensor causes lower oxygen readings.
The processed CTDO dissolved oxygen cannot be expected to have the same absolute
accuracy throughout the water column as the rosette oxygens, although there
have been many instances where individual bottle oxygens have been shown to
be in error by the CTD oxygen trace. ODF therefore recommends that investigators
use bottle oxygen concentrations for absolute values. However, the continuous
oxygen profile is of great value in its detailed structure.
2.1.3.3 Additional Processing
A software filter was used if needed to remove larger conductivity, temperature,
or oxygen spiking problems. Less than 0.5% of the time-series data in the very
few noisy casts was affected. The downcast portion of each time-series was then
reaveraged into 2-decibar pressure intervals. A ship-roll filter was applied
to disallow pressure reversals. Density inversions which still remain in high-gradient
regions sometimes cannot be accounted for by a mis-match of pressure, temperature
and conductivity sensor response. Detailed examination of the raw data typically
shows significant mixing occurring in these areas as a consequence of ship roll.
The best the ship-roll filter can do is to produce a reduction in the number
and/or size of density inversions.
2.2 BEAM ATTENUATION (Gardner, TAMU)
The methods for reducing 25 cm pathlength SeaTech transmissometer data from
the JGOFS North Atlantic Bloom Experiment are described below. In addition,
these same measurements are discussed in more detail in Gardner et al, (1990).
The raw time-series CTD-transmissometer data supplied by SIO ODF group were
decimated to one value every 2 decibars using the first value greater than or
equal to the even decibar value. The decimated transmissometer data were corrected
for aging of the LED light source, index of refraction, temperature and pressure
effects using algorithms supplied by SeaTech. The transmissometer voltages were
converted to percent transmission. Spikes were removed from the data. Beam attenuation
(c) was calculated from the transmission values. The minimum value of (cmin)
was subtracted from all c values to yield (cp) (beam attenuation due to particles
alone). A linear fit was performed between the cp values and the particle mass
concentrations obtained from filtration of water from the Niskin bottles. cp
values were adjusted such that a cp of zero yielded a concentration of zero.
The equation for conversion is Concentration (ug/l) = 1022 *(cp*g) where g=1
for cp>0.1 and g=square root of (cp/0.1) for cp<0.1 This single equation accounts
for a change in the correlation of mass concentration vs. beam attenuation between
surface waters and subsurface waters.
2.3 CTD PROFILES - (Broenkow, MLML)
The MLML CTD/Rosette (Yarbrough et al., 1989) was used to make profiles of conductivity,
temperature, dissolved oxygen, beam attenuation and in situ fluorescence. Conductivity
was measured with a Sea-Bird conductivity cell and MLML pump, temperature with
a platinum thermometer (tau = 0.3 sec) and pressure with a Digiquartz transducer.
Data were digitized at 0.8 m intervals. Corrections were applied to temperature,
salinity, and pressure using laboratory calibrations done before and after the
cruise. Pressure corrections for the compressibility of the Sea-Bird cell were
applied using the algorithm provided by Sea-Bird Electronics. Corrected data
were compared with salinity and temperature field calibration data provided
by the Scripps CTD group. Scripps corrected CTD data and ours show excellent
agreement. Maximum salinity differences between the SIO and MLML profiles are
about +/- 0.02 S.
The oxygen electrode data were obtained with a Beckman polarograph electrode
modified at MLML to obtain near-membrane temperatures. The data have been corrected
to oxygen concentrations by comparison with titrated calibration samples obtained
during ATLANTIS II 119.5. Most of these calibration samples were analyzed by
MLML personnel, and the RMS difference with Scripps titrations was 3 umole/kg.
Oxygen concentrations were computed from oxygen reduction current via the WHOI
algorithm (Owens and Millard, 1984) using near-membrane temperatures and in
situ pressure. Corrections for membrane porosity changes may be large, and cynicism
is advised when using these data.
2.3.2 Beam Attenuation
The MLML transmissometer is a modified Martek instrument based on the Scripps
Visibility Laboratory design (Petzolf and Austin, 1968). Beam attenuation is
measured through the folded 1 m path with a Wratten 45 (480 nm) filter and an
IR blocking filter. Calibration is done in the laboratory by adjusting instrument
gain to a transmission reading of 85.5% in dry air. Drift is estimated aboard
ship before and after each cast by diligent cleaning of the windows using alcohol.
2.3.3 Fluorescence
The MLML profiling fluorometer uses Variosens electronics (Frungel and Koch,
1980) and produces log-scaled signals. Excitation is via a Xenon flash lamp
and a broad band filter (350-550 nm half power). Fluorescence emission was detected
by silicon diode through a 670 nm (half power) long pass filter. These raw data
are converted to "rescaled fluorescence" units by comparison with extracted
pigment analyses. We provided our own chlorophyll calibrations during ATLANTIS
II 119.5 by fluorometric analysis of acetone extracts of water filtered through
Whatman GF/F (0.7 micron) filters. The "rescaled fluorescence" units are numerically
equivalent to chlorophyll-a concentrations in ug/liter. The term "rescaled fluorescence"
is used to acknowledge the fact that fluorescence and chlorophyll concentrations
may not covary because of variation in quantum yield. The RMS difference between
"rescaled fluorescence" and extracted chlorophyll was 0.27 ug/liter.